Clinical and economic impact of ‘ros1 testing’ strategy compared to a ‘no ros1 testing’ strategy in advanced nsclc in spain

Methods: A joint model decision‑tree and Markov model was developed to determine the cost‑effectiveness of testing ROS1 strategy versus a no‑ROS1 testing strategy in Spain.. According t

Clinical and economic impact

of ‘ROS1-testing’ strategy compared to a ‘no-ROS1-of ‘ROS1-testing’

strategy in advanced NSCLC in Spain

Federico Rojo1, Esther Conde2, Héctor Torres3, Luis Cabezón‑Gutiérrez4, Dolores Bautista5, Inmaculada Ramos6, David Carcedo7,8*, Natalia Arrabal9, J Francisco García9, Raquel Galán9 and Ernest Nadal10

Abstract

Background: Detection of the ROS1 rearrangement is mandatory in patients with advanced or metastatic non‑small

cell lung cancer (NSCLC) to allow targeted therapy with specific inhibitors However, in Spanish clinical practice ROS1

determination is not yet fully widespread The aim of this study is to determine the clinical and economic impact of

sequentially testing ROS1 in addition to EGFR and ALK in Spain.

Methods: A joint model (decision‑tree and Markov model) was developed to determine the cost‑effectiveness of

testing ROS1 strategy versus a no‑ROS1 testing strategy in Spain Distribution of ROS1 techniques, rates of testing, pos‑ itivity, and invalidity of biomarkers included in the analysis (EGFR, ALK, ROS1 and PD‑L1) were based on expert opinion and Lungpath real‑world database Treatment allocation depending on the molecular testing results was defined by

expert opinion For each treatment, a 3‑states Markov model was developed, where progression‑free survival (PFS) and overall survival (OS) curves were parameterized using exponential extrapolations to model transition of patients among health states Only medical direct costs were included (€ 2021) A lifetime horizon was considered and a dis‑ count rate of 3% was applied for both costs and effects Both deterministic and probabilistic sensitivity analyses were performed to address uncertainty

Results: A target population of 8755 patients with advanced NSCLC (non‑squamous or never smokers squamous)

entered the model Over a lifetime horizon, the ROS1 testing scenario produced additional 157.5 life years and 121.3 quality‑adjusted life years (QALYs) compared with no‑ROS1 testing scenario Total direct costs were increased up to

€ 2,244,737 for ROS1 testing scenario The incremental cost‑utility ratio (ICUR) was 18,514 €/QALY Robustness of the

base‑case results were confirmed by the sensitivity analysis

Conclusions: Our study shows that ROS1 testing in addition to EGFR and ALK is a cost‑effective strategy compared to

no‑ROS1 testing, and it generates more than 120 QALYs in Spain over a lifetime horizon Despite the low prevalence of

ROS1 rearrangements in NSCLC patients, the clinical and economic consequences of ROS1 testing should encourage

centers to test all advanced or metastatic NSCLC (non‑squamous and never‑smoker squamous) patients

Keywords: C‑ros oncogene 1, Non‑small cell lung cancer, Molecular testing, Biomarker guided selection, Cost‑

effectiveness analysis

© The Author(s) 2022 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which

permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line

to the material If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder To view a copy of this licence, visit http:// creat iveco mmons org/ licen ses/ by/4 0/ The Creative Commons Public Domain Dedication waiver ( http:// creat iveco mmons org/ publi cdoma in/ zero/1 0/ ) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Background

Lung cancer (LC) has a high incidence rate worldwide and is the main cause of cancer deaths (18.0% of all cancer deaths), so it represents a major health problem [1–4] In

Open Access

*Correspondence: david.carcedo@hygeiaconsulting.com

8 Hygeia Consulting, Barcelona, Spain

Full list of author information is available at the end of the article

Spain, according to the Spanish Society of Medical

Oncol-ogy (SEOM), in 2020 lung cancer was responsible for the

highest number of cancer deaths in Spain, causing 22,930

deaths (20.3% of all cancer deaths) [3]

Non-small cell lung cancer (NSCLC) accounts for 85%

of lung cancer cases and is classified into several

histolog-ical subtypes, of which adenocarcinoma is the most

com-mon (55–60% of LC) [5] In these histological subtypes,

a wide variety of oncogenic driver alterations have been

described, such as the presence of translocations or

rear-rangements of the anaplastic lymphoma kinase (ALK)

gene, mutations in the epidermal growth factor

recep-tor (EGFR) gene, rearrangements of the c-ros oncogene

1 (ROS1) gene, and also the presence of aberrant

expres-sion of programmed death-ligand 1 (PD-L1) [1]

Spe-cifically, the ROS1 gene encodes a receptor with tyrosine

kinase activity that is altered by chromosomal

rearrange-ment in several tumor types, including LC where it can

be detected in approximately 1% of NSCLC patients and

appears to be associated with low tobacco exposure and

adenocarcinoma histology [1 6]

Patients with advanced LC generally have a poor

prog-nosis; however, the advent of targeted therapy directed

to oncogenic genetic alterations has created a new

land-scape, especially in NSCLC treatment, providing

sig-nificant improvements in survival and quality of life [7

8] The growing number of targeted therapies to EGFR

and ALK alterations has resulted in a rapid change in

the prognostic of these subtype of NSCLC patients [9]

In particular, targeted therapy with specific inhibitors of

ROS1 rearrangements in patients with advanced NSCLC

has shown longer overall survival than patients treated

with conventional chemotherapy According to several

studies, long-term disease control exerted by crizotinib in

patients with ROS1 rearrangement is almost double that

the control obtained in patients with ALK alterations [6

10–13] In addition, other drugs, such as entrectinib,

bri-gatinib, lorlatinib and ceritinib, are being studied to treat

patients harboring ROS1-positive cancers [1], but at the

time of the analysis they are not yet available, although

they are at different stages of the approval, pricing and

reimbursement process

In Spain, the SEOM and the Spanish Society of

Pathol-ogy (SEAP) have published a clinical guideline to guide

biomarker testing in patients with advanced NSCLC

[1] According to national and international

recommen-dations for molecular diagnosis in advanced NSCLC

patients, molecular testing of EGFR and BRAF mutations,

ALK and ROS1 rearrangements and PD-L1 expression

are considered mandatory [1 14] ROS1 rearrangement

should be tested in patients with advanced stage

(IIIB-IV) non-squamous NSCLC, regardless of its clinical

characteristics and should not be tested in squamous

cell carcinoma (except in the context of patients with no

or low tobacco exposure and younger than 50 years) [1

7 14] However, although the determination of ROS1 is

mandatory according to guidelines, real-world evidence obtained from Lung Cancer Biomarker Testing Registry

(LungPath) show that ROS1 fusions were not determined

in almost half of the samples of patients with NSCLC (testing rate: 58.1%) [15] According to the Thoracic Tumor Registry (TTR), an observational study also con-ducted in Spanish hospitals (up to the year 2018), showed

even lower ROS1 tests (testing rate [with FISH]: 11.6%])

[16] This low rate of ROS1 testing may be due to the low prevalence of ROS1 rearrangements in patients with

NSCLC that could discourage its determination in some centers, also conditioned by limited diagnostic and/or sampling resources [1 6 15]

Essentially, there are three methodological approaches

to detecting ROS1 rearrangements:

immunohisto-chemistry (IHC), cytogenetic techniques (particularly fluorescent in  situ hybridization [FISH], and molecular techniques such as real-time polymerase chain reaction (RT-PCR) or next-generation sequencing (NGS) [1 17]

To determine ROS1 translocation in clinical specimens,

national and international guidelines recommend IHC

as the screening method and confirmation of positive cases with another orthogonal method (cytogenetic or molecular) like FISH [1 7] FISH is often considered the

gold-standard in the detection of ROS1 rearrangement,

although RT-PCR and NGS (DNA- or RNA-based) also show accurate results in most published studies [18–21]

Based on the clinical implications of ROS1 fusion

detection in NSCLC patients, it is crucial to accurately

identify ROS1 alterations while minimizing response

time [1 22] The importance of testing for other

biomark-ers, such as ALK, has already been quantified in Spain by

Nadal et al [23], however, it has not been quantified for the determination of a less prevalent biomarker such as

ROS1 For this reason, the main objective of this

analy-sis was to quantify the clinical and economic impact of

ROS1 determination in patients with advanced NSCLC

in Spain, comparing a testing ROS1 strategy with sequen-tially testing ROS1 in addition to EGFR and ALK versus a no-testing ROS1 strategy.

Methods

In line with the previous model developed by Nadal

et al [23], a joint model combining a decision-tree with Markov models was developed to determine long-term health results and associated costs of patients with

NSCLC, but in this case by comparing a testing ROS1 strategy by comparison against a no-ROS1 testing

strat-egy in Spain, using Microsoft Excel (Fig. 1)

The decision-tree models comprise the diagnostic

phase, where the sequential determination of EGFR,

ALK, ROS1 and PD-L1 were established In case of a

positive result for any of these biomarkers, the patient

exits the model and receives the corresponding

tar-geted treatment In the model, in case of a negative

result for EGFR, ALK and ROS1 (defined as wild type

[WT] patients), the level of PD-L1 expression is

deter-mined and the result is categorized as Tumor

Propor-tion Score (TPS) ≥ 50% or TPS < 50% This threshold

of PD-L1 expression was defined based on the

indi-cation of pembrolizumab monotherapy for patients

with high PD-L1 expression without oncogenic

altera-tions in EGFR and ALK At some point, the results for

EGFR, ALK and ROS1 can also be invalid, in which case

patients will be direct candidates for re-biopsy In the

excludes only ROS1, so in case of a negative result for

EGFR and ALK, patients are directly considered as

WT patients, and then the level of PD-L1 expression is

determined

Based on the determination results, a specific

treat-ment is assigned (Fig. 1) and patients enter in the

respective Markov model with different long-term

clin-ical and economic outcomes The Markov models are

developed following an area under the curve structure

with three health states: progression-free survival (PFS

state), progressed-disease (PD state), and death state

(absorbent state)

In line with the recommendations by the guidelines

for the evaluation of health technologies in Spain, the

duration of the Markov cycle was 1 month, the time horizon was 20-years (lifetime) and the discount rate for future costs and effects was 3% [24, 25]

The analysis was performed from the perspective of the Spanish National Health System (NHS), so only direct medical costs were considered (expressed in € 2021) The health consequences include life years (LY), progression-free life years (PF-LY) and quality-adjusted life years (QALYs)

The included parameters, the assumptions made as well as the clinical feasibility of the results were vali-dated by a multidisciplinary group of oncologists and pathologists, who are also the authors of this article

Target population

The definition of the target population was similar to the one used in the previous model developed by Nadal

et  al (2021) [23] A hypothetical cohort of patients with advanced or metastatic NSCLC, who were ‘theoretical’ candidates for the molecular diagnosis, was initially esti-mated Therefore, both patients with non-squamous his-tology and those with squamous NSCLC who were never smokers were considered, following the current clinical guidelines for molecular diagnosis in advanced NSCLC [1] The estimation of the target population is shown in Table 1

As shown in Fig. 1, the diagnostic sequence starts

with EGFR, so of the theoretical patients estimated in

Table 1, only those finally tested for EGFR entered the

model

Fig 1 Joint model diagram combining a decision‑tree model with Markov model * ROS1 determined by IHC, FISH, reflex or NGS in ‘ROS1‑testing’

scenario Not determined in ‘no‑ROS1‑testing’ scenario EGFR: epidermal growth factor receptor; ALK: anaplastic lymphoma kinase; ROS1: c‑ros oncogene 1; PD‑L1: programmed death‑ligand 1; pembro: pembrolizumab monotherapy; CT: Chemotherapy; TKI: Tyrosine kinase inhibitors; PFS: progression‑free survival; PD: progression disease

Decision‑tree parameters

All the inputs that are used in the decision-tree

sub-model are listed in Table 2

As shown in Fig. 1, the results of the biomarker

deter-minations can be informative (positive or negative) or

invalid, mainly due to insufficient sample The

positiv-ity rates for EGFR, ALK and ROS1 determinations were

obtained from Lungpath database while the PD-L1

posi-tivity rate (considering TPS > 50% as the threshold for

positivity) was agreed by the expert panel, given that the

PD-L1 positivity rate obtained from Lungpath probably

reflects a mixture of positivity rates with different

thresh-olds depending on the center Invalid rates for each

bio-marker were obtained from the Lungpath database On

the other hand, based on the experience of the experts,

repeating invalid results does not usually give

informa-tive results, so it was assumed that invalid results would

be direct candidates for re-biopsy (considered

success-ful by experts in only 33.3% of cases) When re-biopsy is

unsuccessful, patients receive doublet of chemotherapy if

the molecular diagnosis of EGFR is unknown (due to an

invalid result at the beginning of the sequential determi-nation), or chemo-immunotherapy (the same received by patients with TPS < 50% or unknown PD-L1) if the

inva-lid result was obtained for ALK or ROS1 but the diagno-sis of EGFR is known and negative.

The current distribution of ROS1 determination

tech-niques in Spain included in the model was obtained from

the panel of experts, as the data provided by Lungpath

reflects clinical practice in 2008 and does not correspond with current guideline recommendations where reflex to

FISH is mandatory for ROS1 determination In the base

case, the accuracy of the techniques is not taken into

account, so that the specificity and sensitivity of all ROS1

determination techniques were assumed to be 100% For

this reason, in the base case, the distribution of ROS1

detection techniques only has an impact in terms of costs (not on health outcomes) However, an alternative sce-nario to the base case has been explored where the

accu-racy of the ROS1 determination techniques is considered

The parameter values for IHC were obtained from the values of the IHC clones from the ROSING study [21]

Table 1 Estimated target population

NSCLC Non-small cell lung cancer, EGFR Epidermal growth factor receptor

Table 2 Main decision‑tree inputs

EGFR Epidermal growth factor receptor, ALK Anaplastic lymphoma kinase, ROS1 C-ros oncogene 1, PD-L1 Programmed death-ligand 1, IHC Immunohistochemistry, FISH Fluorescent in situ hybridization, NGS Next-generation sequencing

Invalid results and positivity rate of selected biomarkers

Probability of re‑biopsy

ROS1 determination strategies

and the distribution of clones agreed by the experts (70%

SP384, 30% D4D6) The resulting sensitivity and

specific-ity of IHC was 90.9 and 99.0%, respectively For FISH, the

specificity and sensitivity values agreed by the experts

were 99 and 95%, respectively, and for NGS, the values

of both parameters were assumed to be 100% despite that

in rea-life testing the sensitivity and sensitivity of NGS

would not be 100%

The specific costs of the decision-tree were the cost of

re-biopsy and the costs of the tests used for molecular

diagnosis For the re-biopsy, a cost of € 385.41 was

con-sidered It was calculated by weighting the distribution

of the type of biopsy performed (biopsy, cytology, blood:

65, 30, 5%, respectively) as reported by experts and the

cost of each of the biopsy techniques (€ 555.70, € 58.70,

€ 131.84, respectively) [29] The costs of the tests were

agreed by the expert panel, taking into consideration

the market price: € 70 for IHC; € 110 for FISH; € 455 for

NGS; € 120 for the EGFR test; € 76 for the ALK test; and

€ 70 for the PD-L1 test [29]

As described in Table 3, thirteen treatments and

inclu-sion in clinical trials were included in the analysis, and

for each, costs and long-term outcomes are quantified

using a specific Markov model The expert panel

estab-lished the distribution of all the most common first-line

treatments in Spain for each molecular profile (Table 3)

An alternative scenario to the base case has been

explored considering a potential treatment distribution

scenario where ROS1-positive patients may also receive

entrectinib, as although it is not yet commercially

avail-able for these patients in Spain, experts estimate that

its use will increase in the future and could have some

impact on the model results, unlike other upcoming

ALK-targeted therapies that would have negligible effects

on the results (Table 3)

Depending on the molecular diagnosis result, patients

were assigned to a specific treatment and entered the

respective Markov model developed, in which

effi-cacy and costs associated with each treatment were

considered

Markov model parameters

To establish the transition of the hypothetical cohort

between the health states of the Markov models and

given that the time horizon of the analysis (lifetime) is

longer than the observation periods of the clinical trials,

it is necessary to extrapolate the Kaplan-Meier curves of

PFS and OS to the long term In the absence of

individu-alized data for all treatments to explore different

para-metric distributions, the panel of experts assumed, in

line with the previous model developed, the exponential

models based on the median PFS and OS reported in the

respective studies [23]

Median PFS and OS for ALK targeted therapies

(alec-tinib, crizotinib) were obtained from the recent update of the ALEX study [30], except the median OS in the alec-tinib group was not reached and extrapolation curves were obtained from the alectinib cost-effectiveness model

(data on file) Median PFS and OS for EGFR-targeted

therapies were obtained from FLAURA study (assuming the same efficacy for afatinib as for erlotinib and gefi-tinib) [31, 32] and from ARCHER 1050 study for dac-omitinib [33] The PROFILE 1001 study was used for the median PFS and OS of crizotinib as a targeted therapy for

ROS1 [10], and for the median PFS and OS of entrectinib, the entrectinib cost-effectiveness model (data on file) was used, given that in its STARTRK-2 study, the median OS has not yet been reached For WT patients treated with pembrolizumab in monotherapy, median PFS and OS were obtained from KEYNOTE-024 [34, 35] and for WT patients treated with pembrolizumab in combination and cisplatin + pemetrexed, the medians of the parameters were obtained from the KEYNOTE-189 study (compara-tor and control arm, respectively) [36] For WT patients with TPS < 50%, median PFS and OS for the remaining two treatment strategies considered in the model were obtained from Sandler et al (2006) [37] and IMpower150

Table 3 Distribution of treatments according to molecular

diagnosis

EGFR Epidermal growth factor receptor, ALK Anaplastic lymphoma kinase, ROS1

C-ros oncogene 1, WT Wild-type, TPS Tumour proportion score, Cisp Cisplatin,

Carb Carboplatin, pmtrx Pemetrexed, paclitx Paclitaxel, beva Bevacizumab

a Distribution considered in the alternative scenario in which entrectinib is a treatment alternative in ROS1-positive patients

WT TPS ≥50% Pembrolizumab monotherapy 90%

Cisp+pmtrx+pembrolizumab 60%

Carb+ paclitx

[38] For patients entering a clinical trial, a small

percent-age of ROS1-positive and WT with TPS ≥50% patients,

the experts assumed extrapolation with the longest

medi-ans of the corresponding therapeutic target (medimedi-ans PFS

and OS of crizotinib and pembrolizumab monotherapy,

respectively)

In the alternative scenario considering the accuracy of

the ROS1 determination techniques it was necessary to

establish the clinical consequences of the false positive

(FP) results for ROS1 In this regard, the expert panel

agreed to keep the same assumptions made in the

previ-ous model developed Thus, it was considered that most

patients will have shown progression at the first

follow-up visit and that all of them will have progressed at the

second visit, assuming a median PFS of 2 months

(consid-ering that no patient exceeds 6 months of treatment -stop

rule-) and a median OS of 18 months [23]

The costs of the Markov models included drug

acqui-sition costs (first line and subsequent treatments) [see

Additional file 1] and its associated cost of administration

in case of intravenous drugs (€ 211) [29]

For the acquisition costs, all drug costs are expressed

as the ex-factory price considering the corresponding

deductions according to RDL 08/2010 [39, 40] where

appropriate For drugs where the dose depends on the

patient’s characteristics, the same demographic

char-acteristics of the hypothetical cohort from the previous

model were assumed, that is, a mean body surface area of

1.81 and a mean weight of 72.885 kg [41] The vial sharing

was assumed for intravenous treatment in line also with previous model For entrectinib, which was not yet priced

at the time of analysis, an ex-factory price 10% higher than crizotinib was assumed Clinical trials are assumed

to have no cost to the Spanish NHS

The efficacy of the treatments (in terms of median PFS and median OS) and the costs associated with them are shown in an additional file [see Additional file 1]

Concerning the costs of subsequent treatments admin-istered once patients progress to first-line treatment, only the costs of second-line treatments were considered

in order to simplify the model Both the proportion of patients who would receive active second-line treatment and those who would receive best supportive care (BSC),

as well as the distribution of the most representative sec-ond-line treatments (depending on the first-line received) was established by the experts Median PFS of all subse-quent treatments were obtained from the literature [11,

42–46] The parameters related to the second-line treat-ments are shown in Table 4

Regarding the utility values included, the experts decided to use the same values as those applied in the previously developed model (0.814 for the PFS state and 0.725 and 0.470 for the PD state with and without active treatment, respectively) [23, 49]

Sensitivity analysis

The variables used in the model have some uncertainty

To assess and determine the robustness of the results

Table 4 Definition of subsequent (second‑line) treatments

For the grouping of PbCT/lorlatinib and immunotherapies/doce+nintedanib an arithmetic mean was considered

a expert panel

b [ 47 ]

c [ 48 ]

EGFR Epidermal growth factor receptor, ALK Anaplastic lymphoma kinase, ROS1 C-ros oncogene 1, WT wild-type, Carb Carboplatin; pmtrx: pemetrexed, paclitx

Paclitaxel, beva Bevacizumab, PbCT Platinum-based chemotherapy, doce Docetaxel, BSC Best supportive care

WT

obtained, both deterministic (alternative scenarios to the

base case and univariate analysis) and probabilistic

sensi-tivity analyses were performed

Alternatively to the main analysis (base case), two

scenarios (described throughout the article) were also

explored within the sensitivity analysis:

• Considering a potential scenario in which entrectinib

(not present in the base case scenario) is a treatment

alternative in ROS1 positive patients (Table 3)

• Considering the accuracy of the different techniques

for ROS1 determination, where the specificity and

sensitivity parameters of the different ROS1

determi-nation techniques are included (detailed in 2.3

Deci-sion-tree parameters section)

In the univariate deterministic analysis (one-way

sensi-tivity analysis), some variables of the model were

individ-ually modified, depending on the degree of uncertainty

associated with the variable, by 10% or 20% with respect

to the base case value

In the probabilistic sensitivity analysis (PSA), in line

with recommendations in the literature [50], 1000

simu-lations were run by Monte-Carlo method simultaneously

modifying all parameters with an established

distribu-tion The biomarker prevalence variables, body weight

and body surface area and the probability of re-biopsy in

case of an invalid result, were modified by a normal

dis-tribution, utility values by a beta disdis-tribution, and unit

costs were modified following a gamma distribution

Results

Main analysis (base case)

The results of the base case are reported in Table 5 and

are shown graphically in Fig. 2 (health outcomes) and

Fig. 3 (cost outcomes)

In the defined target population, the strategy of

testing-ROS1 in patients with advanced NSCLC provided a gain

of 121.25 QALYs compared with the no-testing ROS1

strategy over a 20-year time horizon Testing ROS1

strat-egy in these patients also entailed higher costs,

includ-ing those of the tests themselves and the re-biopsies,

but mainly due to the cost of targeted treatments The

comparison of costs and health outcomes through the

incremental cost-utility ratio (ICUR), shows that the

test-ing ROS1 strategy in Spain is cost-effective (€ 18,514/

QALYs), as it was below the cost-effectiveness thresholds

commonly considered in Spain [51, 52]

Sensitivity analysis

The cost and health results of the alternative scenarios

are in line with those of the base case (Table 5) The

mod-ifications made in both analyses from the base case affect

only the costs and health outcomes of the testing ROS1 strategy (no testing ROS1 strategy remained the same).

In the alternative scenario considering a poten-tial future scenario in which entrectinib is available in

ROS1-positive patients, the testing ROS1 vs no-testing ROS1 strategy remains cost-effective, with an ICUR

ratio of € 17,652/QALYs (slightly lower than in the base case) According to the results, although the inclusion of

entrectinib in the treatment of ROS1-positive patients

results in a slight increase in the treatment costs of

test-ing ROS1 strategy compared to the correspondtest-ing base

case strategy (€ 239,199 more € than in base case vs € 1,044,375,352 in base case), it also results in an increase

in QALYs gained (19.47 QALYs more than in base case; 140.72 QALYs in alternative scenario vs 121.25 QALYs in base case)

In the alternative scenario that considers the accuracy

of the techniques the testing ROS1 vs no-testing ROS1

strategy is dominant, generating more QALYs at a lower cost Although in this analysis there is a slight decrease in

QALYs gained from the testing ROS1 strategy compared

to the corresponding base case strategy (23.1 QALYS less than in base case; 98.15 QALYs in alternative scenario vs 121.25 QALYs in base case), there is a more significant decrease in the costs of the strategy compared to the base case (€ 8,257,604 less than in base case vs € 1,048,814,795

in base case), due mainly to the treatment costs

The results of the univariate analysis are represented by

a tornado diagram in Fig. 4, showing how the variations

of each variable analyzed modify the ICUR of the base case (€ 18,514 /QALYs) Discount rate (for both cost and effects), followed by utilities show the greatest impact on the ICUR of the base case

In the PSA, the means obtained from the 1000 simu-lations (€ + 2,209,967 and 18,456 QALYs gained with

Table 5 Base case results: cost‑effectiveness of testing ROS1

strategy vs no‑testing ROS1

ROS1 C-ros oncogene 1, PF Progression-free, LY Life years, QALY Quality-adjusted

life years, ICER Incremental cost-effectiveness ratio, ICUR Incremental cost-utility

ratio

Testing ROS1 No-testing ROS1 Difference

Cost of testing € 2,843,608 € 2,149,256 € + 694,352 Cost of re-biopsy € 72,543 € 45,451 € + 27,093 Cost of treatment € 1,045,898,644 € 1,044,375,352 € + 1,523,292

Total costs € 1,048,814,795 € 1,046,570,058 € + 2,244,737